General FEA Load Modeling Question

[Jieve]

I have a general question about FEA modeling of parts/assemblies subjected to complex loadings. A friend of mine who is a PE and oversees a couple mechanical engineers at his company (they design chemical processing plants) and I were having a discussion about general design of machine parts to avoid failure. I am a mechanical engineer currently working in technical education, and am designing a small machine used to demonstrate machine vibration phenomena (resonance, single plane balance, critical speeds, etc). I have the luxury to work in an environment where I have little time pressure in comparison to industry, and am able to research in-depth different methods of solving and modeling different engineering design problems. He is very familiar with the project that I am working on, and we began discussing the extent to which detail should be incorporated into the FEA models to verify designs. In my case, this machine is basically a plate mounted on springs, on which are mounted two motor driven shafts with imbalance discs. The plate is free to vibrate in all directions … A detailed analysis would require choosing some operating points where the loads and accelerations would be highest, setting up free body diagrams, use the 3-2-1 method to constrain the parts then apply all balanced forces to the mechanism. However, since there isn’t much symmetry and there are 3 rotational and 3 translational degrees of freedom, there are products of inertia, angular accels and velocities and gyroscopic type forces to consider making the hand calcs time consuming. I did some approximate calculations using forces a little higher than max at the load points, but honestly there are a large number of different possibilities and it seems a bit overwhelming trying to figure out which loading scenarios should be used that would actually result in the highest stresses. Instead, as one example, since I am interested in the bending stiffness of the plate, I reasoned that if I were to fix one end on the plate and apply the maximum loads at the load points that will be seen in service, if I could test different rib setups and keep the deflection below a certain minimum, then as long as the plate can handle that unrealistic scenario, it will remain rigid enough in service. But then the question becomes, if I know the part will deflect 4mm under this over-loading, but in practice it needs to be under 1mm, how do I know this is enough? I’m guessing that for many designs, there are codes that regulate this (based on testing procedures). But for non-standard parts, I’m guessing a test needs to be developed, or what do you think? This friend of mine agreed that he often has young engineers overcomplicate the problems and spend way too much time giving him an answer to 6 decimal places when the loads have to be estimated anyway. What I’m curious about is, how common is this type of scenario (applying unrealistic but simplified loadings that are more than would ever be encountered in practice) to simplify the problem? In my more academic environment, I have more time to focus on the details and work through different scenarios, but in industry it’s more like, you learn as much as you need to get the job done then move on. Curious what your comments are.

 

[flash3780]

Whew... that was some post. Let me try to sum up: You're trying to design a resonance machine and you're having trouble determining the maximum deflection under a given load. You've estimated a conservative load somehow... and that leads to excessive deflection.

Well, if you're dealing with resonance, obviously static loading will be insufficient to predict the behavior; you need to do a dynamic analysis. A finite element analysis can be done, but damping is very difficult to know... which is a problem since your device is meant to resonate. Most companies who try and model forced vibration do so with damping data available from similar hardware.

Honestly, for a simple device, I'd suggest that you build it and test it. You can adjust the amount of rotating imbalance until the maximum amplitude is acceptable. I wish I could give a more satisfying answer, but damping is a really tough thing to accurately quantify.

 

[rb1957]

when do you want to spend your money ?

1) spend a real lot on material ... over-size it so much that there "clearly" isn't a vibration problem;
2) spend a bunch on material, over-size it and a little on testing (to be sure, to give yourself a database to extraploate from ...);
3) spend a bunch on analysis, developing design methods, hrs and hrs of FEA and testing (but you'll have a weight efficient design (if that's worth anything to you) and a whole lot of knowledge of your product's workplace ... a spin-off would be the technical reports you'd be able to publish);
4) save your money and spend it on warantee claims.

I often say at work that i'm not interested in how strong something is, only that it's strong enough. For us the key design feature is ease (and cheapness) of manufacturing; a pound or two of Al is weight well wasted (especially if it'd take a tonne of analysis and a sizeable portion of the US debt to take it out).

 

[Jieve]

Guys, thanks for your input.

@flash - Would a quasi-static analysis at specific points in time (i.e. when the external & inertial loads are greatest) not be enough to get an idea of the stress distribution in the part? Concurrently with the drawing of the parts I also developed a number of relatively elaborate matlab/simulink simulations to make sure this thing was going to do what I wanted it to, for spring damping I think I hovered around a damping ratio of 5%. However I ran simulations with damping ratios between 0.5-10% just to make sure and I was satisfied with the results.

@rb1957 - this was exactly where I was going with my question. My PE friend was saying he takes your "will it do the job" approach as well, I'm guessing tending toward 2, even 1 on rare occasion, when he knows that it's going to be his ass on the line if something goes wrong, especially in an industry where catastrophic failure could mean lives. My impression is that over-designing and spending a little extra on the material (especially for a small project like this) makes up for the time and money spent on analysis to get a very weight efficient design. I tend to like to optimize things, which makes it very difficult for me to finally step back say, "ok this is good enough" (I think this is a fundamental personality characteristic of mine). But as this project is also a learning experience for me, I would like to hone my optimization skills as well. But from a business standpoint, it's ultimately about cost efficiency - money in vs. money out, which is where I was going with the simplified analysis for over-design vs. very elaborate, time-consuming but optimized design of the vibrating baseplate.

Matt Stracker

 

[flash3780]

"Would a quasi-static analysis at specific points in time (i.e. when the external & inertial loads are greatest) not be enough to get an idea of the stress distribution in the part?"

I haven't seen your design or analysis, but without knowing the damping ratio, the multiplication factor can vary significantly. Making your machine adjustable, such that the damping or rotating imbalance could be varied until the desired deflection at resonance is achieved is a reasonable, scientific way of both achieving your design goal (a machine which demonstrates resonance... and doesn't blow up) and determining the damping in your system. Once you've determined the proper amount of damping adjustment or imbalance force that's required, you may be able to remove the adjustments from the production design.